Monitoring vegetation degradation using remote sensing and machine learning over India – a multi-sensor, multi-temporal and multi-scale approach

Frontiers in Forests and Global Change, Journal Year: 2024, Volume and Issue: 7

Published: June 6, 2024

Vegetation cover degradation is often a complex phenomenon, exhibiting strong correlation with climatic variation and anthropogenic actions. Conservation of biodiversity important because millions people are directly indirectly dependent on vegetation (forest crop) its associated secondary products. United Nations Sustainable Development Goals (SDGs) propose to quantify the proportion as total land area all countries. Satellite images form one main sources accurate information capture fine seasonal changes so that long-term can be assessed accurately. In present study, Multi-Sensor, Multi-Temporal Multi-Scale (MMM) approach was used estimate vulnerability degradation. Open source Cloud computing system Google Earth Engine (GEE) systematically monitor evaluate potential multiple satellite data variable spatial resolutions. Hotspots were demarcated using machine learning techniques identify greening browning effect coarse resolution Normalized Difference Index (NDVI) MODIS. Rainfall datasets Climate Hazards Group InfraRed Precipitation Station (CHIRPS) for period 2000–2022 also find rainfall anomaly in region. Furthermore, hotspot areas identified high-resolution major based analysis understand verify cause change whether or nature. This study several State/Central Government user departments, Universities, NGOs lay out managerial plans protection vegetation/forests India.

Language: Английский

---

UAV-Based Classification of Intercropped Forage Cactus: A Comparison of RGB and Multispectral Sample Spaces Using Machine Learning in an Irrigated Area

AgriEngineering, Journal Year: 2024, Volume and Issue: 6(1), P. 509 - 525

Published: Feb. 23, 2024

Precision agriculture requires accurate methods for classifying crops and soil cover in agricultural production areas. The study aims to evaluate three machine learning-based classifiers identify intercropped forage cactus cultivation irrigated areas using Unmanned Aerial Vehicles (UAV). It conducted a comparative analysis between multispectral visible Red-Green-Blue (RGB) sampling, followed by the efficiency of Gaussian Mixture Model (GMM), K-Nearest Neighbors (KNN), Random Forest (RF) algorithms. classification targets included exposed soil, mulching cover, developed undeveloped cactus, moringa, gliricidia Brazilian semiarid. results indicated that KNN RF algorithms outperformed other methods, showing no significant differences according kappa index both Multispectral RGB sample spaces. In contrast, GMM showed lower performance, with values 0.82 0.78, compared 0.86 0.82, 0.82. performed well, individual accuracy rates above 85% Overall, algorithm demonstrated superiority space, whereas excelled space. Even better performance images, learning applied samples produced promising crop classification.

Language: Английский

---

Can Different Cultivars of Panicum maximum Be Identified Using a VIS/NIR Sensor and Machine Learning?

AgriEngineering, Journal Year: 2024, Volume and Issue: 6(4), P. 3739 - 3751

Published: Oct. 16, 2024

Panicum maximum cultivars have distinct characteristics, especially morphological ones related to the leaf structure and coloration, there may be differences in spectral behavior captured by sensors. These can used classification using machine learning (ML) algorithms differentiate biodiversity within same species. The objectives of this study were identify ML models able P. determine which is best input for these whether reducing sample size improves response algorithms. experiment was carried out at experimental area Forage Sector School Farm belonging Federal University Mato Grosso do Sul (UFMS). samples Massai, Mombaça, Tamani, Quênia, Zuri collected from plots field. Analysis on 120 a VIS/NIR hyperspectral sensor. After obtaining data separating them into bands, submitted analysis classify based variables. tested artificial neural networks (ANNs), REPTree J48 decision trees, random forest (RF), support vector (SVM). A logistic regression (LR) as traditional method. Two evaluated algorithms: entire spectrum band provided sensor (ALL) another configuration calculated bands. reflectances showed different behavior, green NIR regions. RL ANN all information are accurately cultivars, reaching accuracies above 70 CC 0.6 kappa F-score. reflectance powerful tool low-cost, non-destructive, high-performance distinguish cultivars. Here, we achieved better model accuracy only 40 samples. In present study, tree proved good performance regardless used, makes it strategic forage cultivar studies smaller or larger datasets.

Language: Английский

---

Addressing Constraint Coupling and Autonomous Decision-Making Challenges: An Analysis of Large-Scale UAV Trajectory-Planning Techniques

Drones, Journal Year: 2024, Volume and Issue: 8(10), P. 530 - 530

Published: Sept. 28, 2024

With the increase in UAV scale and mission diversity, trajectory planning systems faces more complex constraints, which are often conflicting strongly coupled, placing higher demands on real-time response capabilities of system. At same time, conflicts strong coupling pose challenges autonomous decision-making capability system, affecting accuracy efficiency system environments. However, recent research advances addressing these issues have not been fully summarized. An in-depth exploration constraint handling techniques will be great significance to development large-scale systems. Therefore, this paper aims provide a comprehensive overview topic. Firstly, functions application scenarios introduced classified detail according method, realization function presence or absence constraints. Then, described detail, focusing priority ranking constraints principles their fusion transformation methods. importance is depth, related dynamic adjustment algorithms introduced. Finally, future directions outlooked, providing references for fields clustering cooperative flight.

Language: Английский

---

Multispectral UAV-Based Disease Identification Using Vegetation Indices for Maize Hybrids

Agriculture, Journal Year: 2024, Volume and Issue: 14(11), P. 2002 - 2002

Published: Nov. 7, 2024

In the future, cultivation of maize will become more and prominent. As world’s demand for food animal feeding increases, remote sensing technologies (RS technologies), especially unmanned aerial vehicles (UAVs), are developing more, usability cameras (Multispectral-MS) installed on them is increasing, plant disease detection severity observations. present research, two different hybrids, P9025 sweet corn Dessert R78 (CS hybrid), were employed. Four treatments performed with three doses (low, medium, high dosage) infection smut fungus (Ustilago maydis [DC] Corda). The fields monitored times after inoculation—20 DAI (days inoculation) 27 DAI. orthomosaics created in WebODM 2.5.2 software study included five vegetation indices (NDVI [Normalized Difference Vegetation Index], GNDVI [Green Normalized NDRE Red Edge], LCI [Leaf Chlorophyll Index] ENDVI [Enhanced Index]) further analysis QGIS. gathered data analyzed using R-based Jamovi 2.6.13 statistical methods. case hybrid, we obtained promising results, as follows: NDVI values CS 0 significantly higher than high-dosed 10.000 a mean difference 0.05422 *** p value 4.43 × 10−5 value, suggesting differences all levels infection. Furthermore, investigated correlations (VI) R78, where showed correlations. had strong correlation (r = 0.83), medium 0.56) weak 0.419). There was also between GNDVI, r 0.836. coefficients CCoeff. 0.716. For hybrid separation analyses, useful results well.

Language: Английский

---

Differentiation of Soybean Genotypes Concerning Seed Physiological Quality Using Hyperspectral Bands

AgriEngineering, Journal Year: 2024, Volume and Issue: 6(4), P. 4752 - 4765

Published: Dec. 9, 2024

The use of summarized spectral data in bands obtained by hyperspectral sensors can make it possible to obtain biochemical information about seeds and, thus, relate the results seed viability and vigor. Thus, hypothesis this work is based on possibility obtaining physiological quality through distinguishing lots regarding their wavelengths. objective was then evaluate differentiating soybean genotypes using data. experiment conducted during 2021/2022 harvest at Federal University Mato Grosso do Sul a randomized block design with four replicates 10 F3 populations (G1, G8, G12, G15, G19, G21, G24, G27, G31, G36). After maturation each genotype, were harvested from central rows plot, which consisted five one-meter rows. Seed samples experimental unit placed Petri dish collect Readings performed laboratory temperature 26 °C two 60 W halogen lamps as light source, positioned 15 cm between sensor sample. used Ocean Optics (Florida, USA) model STS-VIS-L-50-400-SMA, captured reflectance sample wavelengths 450 824 nm. readings sensor, subjected tests for water content, germination, first germination count, electrical conductivity, tetrazolium. an analysis variance means compared Scott–Knott test 5% probability, analyzed R software version 4.2.3 (Auckland, New Zealand). principal component (PCA) associated K-means algorithm form groups according similarity distinction genetic materials. formation these groups, curve graphs constructed genotype that formed. be differentiated bands. bands, therefore, provide important seeds. Through observation specific differentiate terms quality, complementing and/or replacing traditional fast, accurate, non-destructive way, reducing time investment spent found study are promising, further research needed future studies other species genotypes. interval 649 nm main spectrum band contributed differentiation superior inferior quality.

Language: Английский

---